Phil 3.30.16

7:00 – 3:30 VTX

  • So I was starting The spreading of misinformation online, but it was discussing more of the same. This feels a lot like saturation. My thoughts are coalescing around the idea of the difference between trusted and trustworthy interactions in computer-mediated systems. The anonymous citizen journalism concept becomes a unifying thought experiment that can be used to show the potential strengths and weaknesses of particular concepts.
  • The last piece I think I need is what is trust from a developmental perspective. The initial google scholar search of “trust development” didn’t bring up exactly what I want (object permanence maybe?), but it did provide this: Effects of four computer-mediated communications channels on trust development. The citations provided this: The mechanics of trust: A framework for research and design In International Journal of Human – Computer Studies 2005 62(3):381-422. This one seems different enough to look through carefully.
  • Ok, I think I found what I’m looking for: The ‘like me’ framework for recognizing and becoming an intentional agent. I think I’ll read The Mechanics of Trust first, them ‘like me’ second.
  • Starting The mechanics of trust: A framework for research and design.
    • It does seem to be focused on how effectively a system transmits(?) cues that support well-placed trust. I think that we tend to confuse the trust we place in the channel vs the trust we place in the entity at the other end of the channel. And these lines are not clearly drawn:
      • In IR, we trust that the search engine is providing us with the relevant documents we seek. People trust Google more than Bing because the results are more pertinent. Does this trust carry over into the documents retrieved? Probably, though I can’t find a study that does this. (It would be pretty easy to do with the Google Custom Search Engine API + noise)
      • In GPS the trust in the system is very high, even though it is synthesizing information from retrieved and processed sources (maps, DTED, etc) that could in turn be wrong. Here though, the entity we are interacting with is clearly the GPS, not the mapmakers.
      • Skype, on the other hand is essentially transparent when it’s working right. And that ‘working right’ is a kind of conditional trust in the system that has no effect on out evaluation of the trustworthiness of the person that we are interacting with at the other end of the channel.
      • So what does that mean in the context of our imaginary citizen journalists?
        • They are anonymized. We have no names. We probably don’t even have the exact words as written. These are the same issues that newspapers face when dealing with anonymous sources. And in this case, it’s reasonable to assume that the newspaper is the entity that is attempting to get us to place our trust in it.
          • Reporters as proxies
          • Additional perspectives – images, videos etc.
          • Stories that match reader’s experiences, so that trust can be evaluated.
          • What else?
    • One of the cited papers is What is Trust? A Conceptual Analysis and An Interdisciplinary Model. Quickly scanning through it, I found this on page 830-831: Garfinkel found in natural experiments that people don’t trust others when things “go weird,” that is, when they face inexplicable, abnormal situations. For example, one subject told the experimenter he had a flat tire on the way to work. The experimenter responded, “What do you mean, you had a flat tire?” The subject replied, in a hostile way, “What do you mean? What do you mean? A flat tire is a flat tire. That is what I meant. Nothing special. What a crazy question!” At this point, trust between them broke down because the illogical question produced an abnormal situation. 
      • I think that this is core. Trust is tied to normalicy, and probably builds out from there. 
  • Prepping for the sprint planning session.
  • As far as the OMG work, I think the following
    • Set up version controlled system for Google CSE keys and url exclude lists, including a way to submit an url for inclusion in an exclusion list.
    • Add PDF parsing and storing to Crawl Service
    • Add MSWord parsing and storing to Crawl Service
    • Add MSExcel parsing and storing to CrawlService
    • Add backlink calculation and storing to CrawlService – this is looking like a good way to increase pertinence within a return, particularly with respect to the matched-name wrong-person condition.
    For the machine learning work
    • Get DB up, accessible and on a backup schedule
    • Set up deployment infrastructure for Rating App.
    • Small scale test of Rating App, with refinement and development of manual
    • Accumulate corpus
    • Test corpus in WEKA
      • Translator from DB to WEKA format
      • Construction of training data sets
      • Tests and evaluations
      • Report
    As far as my research, it’s more vague, so I’m just going to free-associate a bit here.
    First, I just need to write up the proposal, and since that’s where my head is at right now, it’s hard to come up with specifics. One of the overall goals is to build a search result interface that ‘nudges’ users from bubble patterns into star patterns.
    Secondly, it’s my current belief is that this interface could be along the lines of the word cloud plus slider display interface I’ve discussed with you before. On the back end, there’s a topic extraction/document classification system that builds a graph database that is used for:
    • In my case, placing the search results in a context of discussion vs information (DvI) along the axis’ defined by the topics in the search results. The user can select a topic (which then shows the DvI graphs and where the current search falls on those spectrums). Once a topic has been selected, the user can adjust the weights on subsequent topics, causing the result list to reorder and the position on the DvI graph to move.
    • In EIT’s case (1) predictions and alerts and (2) for the user interface [and I think this can be pitched as the gamified display]. For example, I think there are many cases where conditions for making a judgment (medical best practices or behavior related) may be ambiguous. Using such an interface could allow a user to explore and resolve such ambiguity. The nice thing is that in the EIT case, the data is (potentially) more structured and granular, allowing a more fluid analysis (e.g. a bad manager indirectly affecting performance or combined conditions such as opiate addiction + newborn).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: